weka segmentation in batch mode

[William Menegas]

Hello,

I was wondering if anyone could give me some advice on using "Weka
Segmentation" in headless mode or in a normal session of fiji in batch
mode. I am using:

...
run("Trainable Weka Segmentation");
call("trainableSegmentation.Weka_Segmentation.loadClassifier",
"/path/classifier.model");
call("trainableSegmentation.Weka_Segmentation.getProbability");
selectWindow("Probability maps");
...

The problem is that fiji doesn't seem to want to wait for the
classification. I tried inserting "wait(100000)" before
"selectWindow("Probability maps"), but it didn't help and I don't think
there was even a wait.

Any help would be much appreciated!
Will

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Aryeh Weiss]

On 3/21/14, 7:52 AM, William Menegas wrote:

> Hello,
>
> I was wondering if anyone could give me some advice on using "Weka
> Segmentation" in headless mode or in a normal session of fiji in batch
> mode. I am using:
>
> ...
> run("Trainable Weka Segmentation");
> call("trainableSegmentation.Weka_Segmentation.loadClassifier",
> "/path/classifier.model");
> call("trainableSegmentation.Weka_Segmentation.getProbability");
> selectWindow("Probability maps");
> ...
>
> The problem is that fiji doesn't seem to want to wait for the
> classification. I tried inserting "wait(100000)" before
> "selectWindow("Probability maps"), but it didn't help and I don't think
> there was even a wait.
>
> Any help would be much appreciated!
> Will
>
> --
> ImageJ mailing list:http://imagej.nih.gov/ij/list.html
>

I had a similar problem with the EDF plugin. Here is a quote form that post:

> The problem I have is that the macro does not wait for EDF to finish,
> and EDF can take a while.
> My solution to this is based on the fact that EDF creates a window
> called "Output". so I did the following:
>
> initTime = getTime();
> while ( !isOpen("Output") ) {
>     elapsedTime = getTime() - initTime;
>     if (elapsedTime%10000 == 0) {
>         print(elapsedTime/1000, " seconds elapsed");
>         wait(10);
>     }
>
> which works (as  long as there is no other window called "Output"). I
> print an occasional reminder that the macro is still alive within this
> wait loop.

I posted this as a question, wondering if there is a better way to do
it. I did not receive any reply, so perhaps in a macro this is the way
to go. You should replace "Output" with "Probability maps".

--aryeh

--
Aryeh Weiss
Faculty of Engineering
Bar Ilan University
Ramat Gan 52900 Israel

Ph:  972-3-5317638
FAX: [phone]


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Ignacio Arganda-Carreras]

Hello Aryeh and Will,

For the classification task I think it's a bit risky to use a delay anyway
because you don't know how long the classification will take. Have you
tried using a script instead of a macro? That allows you to have much more
control over the images and the commands are executed as in a regular
program, so you won't need any wait call.

Have a look at:

http://fiji.sc/Scripting_the_Trainable_Segmentation

A simple Beanshell script equivalent to your macro code would be:

import trainableSegmentation.*;

 // input train image
 input  = IJ.openImage( "input-grayscale-or-color-image.tif" );
 // create Weka Segmentation object
 segmentator = new WekaSegmentation( input );
 // load classifier from file
 segmentator.loadClassifier( "my-cool-trained-classifier.model" );

 // apply classifier to current training image and get label result
 // (set parameter to true to get probabilities)
 segmentator.applyClassifier( false );
 // get result (float image)
 result = segmentator.getClassifiedImage();
Let me know if you have questions about it!

ignacio



On Fri, Mar 21, 2014 at 7:11 AM, Aryeh Weiss <[hidden email]> wrote:

> On 3/21/14, 7:52 AM, William Menegas wrote:
>
>> Hello,
>>
>> I was wondering if anyone could give me some advice on using "Weka
>> Segmentation" in headless mode or in a normal session of fiji in batch
>> mode. I am using:
>>
>> ...
>> run("Trainable Weka Segmentation");
>> call("trainableSegmentation.Weka_Segmentation.loadClassifier",
>> "/path/classifier.model");
>> call("trainableSegmentation.Weka_Segmentation.getProbability");
>> selectWindow("Probability maps");
>> ...
>>
>> The problem is that fiji doesn't seem to want to wait for the
>> classification. I tried inserting "wait(100000)" before
>> "selectWindow("Probability maps"), but it didn't help and I don't think
>> there was even a wait.
>>
>> Any help would be much appreciated!
>> Will
>>
>> --
>> ImageJ mailing list:http://imagej.nih.gov/ij/list.html
>>
>>  I had a similar problem with the EDF plugin. Here is a quote form that
> post:
>
>> The problem I have is that the macro does not wait for EDF to finish, and
>> EDF can take a while.
>> My solution to this is based on the fact that EDF creates a window called
>> "Output". so I did the following:
>>
>> initTime = getTime();
>> while ( !isOpen("Output") ) {
>>     elapsedTime = getTime() - initTime;
>>     if (elapsedTime%10000 == 0) {
>>         print(elapsedTime/1000, " seconds elapsed");
>>         wait(10);
>>     }
>>
>> which works (as  long as there is no other window called "Output"). I
>> print an occasional reminder that the macro is still alive within this wait
>> loop.
>>
>
> I posted this as a question, wondering if there is a better way to do it.
> I did not receive any reply, so perhaps in a macro this is the way to go.
> You should replace "Output" with "Probability maps".
>
> --aryeh
>
> --
> Aryeh Weiss
> Faculty of Engineering
> Bar Ilan University
> Ramat Gan 52900 Israel
>
> Ph:  972-3-5317638
> FAX: [phone]
>
>
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> Ignacio Arganda-Carreras, Ph.D.
> Seung's lab, 46-5065
> Department of Brain and Cognitive Sciences
> Massachusetts Institute of Technology
> 43 Vassar St.
> Cambridge, MA 02139
> USA
>
> Phone: [phone]
> Website: <http://imagej.nih.gov/ij/list.html>
> http://bioweb.cnb.csic.es/~iarganda/index_EN.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Larry Lindsey]

Hi List,

I ran into a similar problem a little while ago. There's a plugin (full
disclosure: I wrote it) for the 2D case [1]. It's intended to run on a
networked cluster, but it would be simple to modify it for single computers
as well. Let me know if that sounds like it would be useful.

Larry

1: http://fiji.sc/FijiArchipelago#Batch_Weka_Segmentation


On Fri, Mar 21, 2014 at 3:56 AM, Ignacio Arganda-Carreras <
[hidden email]> wrote:

> Hello Aryeh and Will,
>
> For the classification task I think it's a bit risky to use a delay anyway
> because you don't know how long the classification will take. Have you
> tried using a script instead of a macro? That allows you to have much more
> control over the images and the commands are executed as in a regular
> program, so you won't need any wait call.
>
> Have a look at:
>
> http://fiji.sc/Scripting_the_Trainable_Segmentation
>
> A simple Beanshell script equivalent to your macro code would be:
>
> import trainableSegmentation.*;
>
>  // input train image
>  input  = IJ.openImage( "input-grayscale-or-color-image.tif" );
>  // create Weka Segmentation object
>  segmentator = new WekaSegmentation( input );
>  // load classifier from file
>  segmentator.loadClassifier( "my-cool-trained-classifier.model" );
>
>  // apply classifier to current training image and get label result
>  // (set parameter to true to get probabilities)
>  segmentator.applyClassifier( false );
>  // get result (float image)
>  result = segmentator.getClassifiedImage();
> Let me know if you have questions about it!
>
> ignacio
>
>
>
> On Fri, Mar 21, 2014 at 7:11 AM, Aryeh Weiss <[hidden email]> wrote:
>
> > On 3/21/14, 7:52 AM, William Menegas wrote:
> >
> >> Hello,
> >>
> >> I was wondering if anyone could give me some advice on using "Weka
> >> Segmentation" in headless mode or in a normal session of fiji in batch
> >> mode. I am using:
> >>
> >> ...
> >> run("Trainable Weka Segmentation");
> >> call("trainableSegmentation.Weka_Segmentation.loadClassifier",
> >> "/path/classifier.model");
> >> call("trainableSegmentation.Weka_Segmentation.getProbability");
> >> selectWindow("Probability maps");
> >> ...
> >>
> >> The problem is that fiji doesn't seem to want to wait for the
> >> classification. I tried inserting "wait(100000)" before
> >> "selectWindow("Probability maps"), but it didn't help and I don't think
> >> there was even a wait.
> >>
> >> Any help would be much appreciated!
> >> Will
> >>
> >> --
> >> ImageJ mailing list:http://imagej.nih.gov/ij/list.html
> >>
> >>  I had a similar problem with the EDF plugin. Here is a quote form that
> > post:
> >
> >> The problem I have is that the macro does not wait for EDF to finish,
> and
> >> EDF can take a while.
> >> My solution to this is based on the fact that EDF creates a window
> called
> >> "Output". so I did the following:
> >>
> >> initTime = getTime();
> >> while ( !isOpen("Output") ) {
> >>     elapsedTime = getTime() - initTime;
> >>     if (elapsedTime%10000 == 0) {
> >>         print(elapsedTime/1000, " seconds elapsed");
> >>         wait(10);
> >>     }
> >>
> >> which works (as  long as there is no other window called "Output"). I
> >> print an occasional reminder that the macro is still alive within this
> wait
> >> loop.
> >>
> >
> > I posted this as a question, wondering if there is a better way to do it.
> > I did not receive any reply, so perhaps in a macro this is the way to go.
> > You should replace "Output" with "Probability maps".
> >
> > --aryeh
> >
> > --
> > Aryeh Weiss
> > Faculty of Engineering
> > Bar Ilan University
> > Ramat Gan 52900 Israel
> >
> > Ph:  972-3-5317638
> > FAX: [phone]
> >
> >
> >
> > --
> > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
> >
> > --
> > Ignacio Arganda-Carreras, Ph.D.
> > Seung's lab, 46-5065
> > Department of Brain and Cognitive Sciences
> > Massachusetts Institute of Technology
> > 43 Vassar St.
> > Cambridge, MA 02139
> > USA
> >
> > Phone: [phone]
> > Website: <http://imagej.nih.gov/ij/list.html>
> > http://bioweb.cnb.csic.es/~iarganda/index_EN.html
> >
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[William Menegas]

Thanks for all the suggestions!

I'll try the 'quick fix' which is based on Aryeh's suggestion:

call("trainableSegmentation.Weka_Segmentation.getProbability");
waiting = 0;
while (waiting == 0){
if (isOpen("Probability maps") == true)
waiting = 1;
else
wait(10000);
}
selectWindow("Probability maps");

If that doesn't work, I'll try rewriting everything to run as a beanshell
script, or just the weka part as a beanshell script - as Ignacio suggested.

And I was planning on running this on a cluster, so I'll check out Larry's
plugin as well...

Thanks again everybody!
Will


On Fri, Mar 21, 2014 at 12:14 PM, Larry Lindsey <[hidden email]>wrote:

> Hi List,
>
> I ran into a similar problem a little while ago. There's a plugin (full
> disclosure: I wrote it) for the 2D case [1]. It's intended to run on a
> networked cluster, but it would be simple to modify it for single computers
> as well. Let me know if that sounds like it would be useful.
>
> Larry
>
> 1: http://fiji.sc/FijiArchipelago#Batch_Weka_Segmentation
>
>
> On Fri, Mar 21, 2014 at 3:56 AM, Ignacio Arganda-Carreras <
> [hidden email]> wrote:
>
> > Hello Aryeh and Will,
> >
> > For the classification task I think it's a bit risky to use a delay
> anyway
> > because you don't know how long the classification will take. Have you
> > tried using a script instead of a macro? That allows you to have much
> more
> > control over the images and the commands are executed as in a regular
> > program, so you won't need any wait call.
> >
> > Have a look at:
> >
> > http://fiji.sc/Scripting_the_Trainable_Segmentation
> >
> > A simple Beanshell script equivalent to your macro code would be:
> >
> > import trainableSegmentation.*;
> >
> >  // input train image
> >  input  = IJ.openImage( "input-grayscale-or-color-image.tif" );
> >  // create Weka Segmentation object
> >  segmentator = new WekaSegmentation( input );
> >  // load classifier from file
> >  segmentator.loadClassifier( "my-cool-trained-classifier.model" );
> >
> >  // apply classifier to current training image and get label result
> >  // (set parameter to true to get probabilities)
> >  segmentator.applyClassifier( false );
> >  // get result (float image)
> >  result = segmentator.getClassifiedImage();
> > Let me know if you have questions about it!
> >
> > ignacio
> >
> >
> >
> > On Fri, Mar 21, 2014 at 7:11 AM, Aryeh Weiss <[hidden email]>
> wrote:
> >
> > > On 3/21/14, 7:52 AM, William Menegas wrote:
> > >
> > >> Hello,
> > >>
> > >> I was wondering if anyone could give me some advice on using "Weka
> > >> Segmentation" in headless mode or in a normal session of fiji in batch
> > >> mode. I am using:
> > >>
> > >> ...
> > >> run("Trainable Weka Segmentation");
> > >> call("trainableSegmentation.Weka_Segmentation.loadClassifier",
> > >> "/path/classifier.model");
> > >> call("trainableSegmentation.Weka_Segmentation.getProbability");
> > >> selectWindow("Probability maps");
> > >> ...
> > >>
> > >> The problem is that fiji doesn't seem to want to wait for the
> > >> classification. I tried inserting "wait(100000)" before
> > >> "selectWindow("Probability maps"), but it didn't help and I don't
> think
> > >> there was even a wait.
> > >>
> > >> Any help would be much appreciated!
> > >> Will
> > >>
> > >> --
> > >> ImageJ mailing list:http://imagej.nih.gov/ij/list.html
> > >>
> > >>  I had a similar problem with the EDF plugin. Here is a quote form
> that
> > > post:
> > >
> > >> The problem I have is that the macro does not wait for EDF to finish,
> > and
> > >> EDF can take a while.
> > >> My solution to this is based on the fact that EDF creates a window
> > called
> > >> "Output". so I did the following:
> > >>
> > >> initTime = getTime();
> > >> while ( !isOpen("Output") ) {
> > >>     elapsedTime = getTime() - initTime;
> > >>     if (elapsedTime%10000 == 0) {
> > >>         print(elapsedTime/1000, " seconds elapsed");
> > >>         wait(10);
> > >>     }
> > >>
> > >> which works (as  long as there is no other window called "Output"). I
> > >> print an occasional reminder that the macro is still alive within this
> > wait
> > >> loop.
> > >>
> > >
> > > I posted this as a question, wondering if there is a better way to do
> it.
> > > I did not receive any reply, so perhaps in a macro this is the way to
> go.
> > > You should replace "Output" with "Probability maps".
> > >
> > > --aryeh
> > >
> > > --
> > > Aryeh Weiss
> > > Faculty of Engineering
> > > Bar Ilan University
> > > Ramat Gan 52900 Israel
> > >
> > > Ph:  972-3-5317638
> > > FAX: [phone]
> > >
> > >
> > >
> > > --
> > > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
> > >
> > > --
> > > Ignacio Arganda-Carreras, Ph.D.
> > > Seung's lab, 46-5065
> > > Department of Brain and Cognitive Sciences
> > > Massachusetts Institute of Technology
> > > 43 Vassar St.
> > > Cambridge, MA 02139
> > > USA
> > >
> > > Phone: [phone]
> > > Website: <http://imagej.nih.gov/ij/list.html>
> > > http://bioweb.cnb.csic.es/~iarganda/index_EN.html
> > >
> >
> > --
> > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
> >
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Aryeh Weiss]

On 3/21/14, 7:00 PM, William Menegas wrote:

> Thanks for all the suggestions!
>
> I'll try the 'quick fix' which is based on Aryeh's suggestion:
>
> call("trainableSegmentation.Weka_Segmentation.getProbability");
> waiting = 0;
> while (waiting == 0){
> if (isOpen("Probability maps") == true)
> waiting = 1;
> else
> wait(10000);
> }
> selectWindow("Probability maps");
>
> If that doesn't work, I'll try rewriting everything to run as a beanshell
> script, or just the weka part as a beanshell script - as Ignacio suggested.

If so, here is a better version. It does not require that elapsedTime be
exactly 10000ms when you test it:

     initTime = getTime();
     oldTime = initTime;
     while (!isOpen("Output")) {
         elapsedTime = getTime() - initTime;
         newTime = getTime() - oldTime;
         if (newTime > 10000) {
             oldTime = getTime();
             newTime = 0;
             print(elapsedTime/1000, " seconds elapsed");
         }

     }
     wait(1000); // to make sure that the window is really ready...

I hope it helps

--aryeh

> And I was planning on running this on a cluster, so I'll check out Larry's
> plugin as well...
>
> Thanks again everybody!
> Will
>
>
> On Fri, Mar 21, 2014 at 12:14 PM, Larry Lindsey <[hidden email]>wrote:
>
>> Hi List,
>>
>> I ran into a similar problem a little while ago. There's a plugin (full
>> disclosure: I wrote it) for the 2D case [1]. It's intended to run on a
>> networked cluster, but it would be simple to modify it for single computers
>> as well. Let me know if that sounds like it would be useful.
>>
>> Larry
>>
>> 1: http://fiji.sc/FijiArchipelago#Batch_Weka_Segmentation
>>
>>
>> On Fri, Mar 21, 2014 at 3:56 AM, Ignacio Arganda-Carreras <
>> [hidden email]> wrote:
>>
>>> Hello Aryeh and Will,
>>>
>>> For the classification task I think it's a bit risky to use a delay
>> anyway
>>> because you don't know how long the classification will take. Have you
>>> tried using a script instead of a macro? That allows you to have much
>> more
>>> control over the images and the commands are executed as in a regular
>>> program, so you won't need any wait call.
>>>
>>> Have a look at:
>>>
>>> http://fiji.sc/Scripting_the_Trainable_Segmentation
>>>
>>> A simple Beanshell script equivalent to your macro code would be:
>>>
>>> import trainableSegmentation.*;
>>>
>>>   // input train image
>>>   input  = IJ.openImage( "input-grayscale-or-color-image.tif" );
>>>   // create Weka Segmentation object
>>>   segmentator = new WekaSegmentation( input );
>>>   // load classifier from file
>>>   segmentator.loadClassifier( "my-cool-trained-classifier.model" );
>>>
>>>   // apply classifier to current training image and get label result
>>>   // (set parameter to true to get probabilities)
>>>   segmentator.applyClassifier( false );
>>>   // get result (float image)
>>>   result = segmentator.getClassifiedImage();
>>> Let me know if you have questions about it!
>>>
>>> ignacio
>>>
>>>
>>>
>>> On Fri, Mar 21, 2014 at 7:11 AM, Aryeh Weiss <[hidden email]>
>> wrote:
>>>> On 3/21/14, 7:52 AM, William Menegas wrote:
>>>>
>>>>> Hello,
>>>>>
>>>>> I was wondering if anyone could give me some advice on using "Weka
>>>>> Segmentation" in headless mode or in a normal session of fiji in batch
>>>>> mode. I am using:
>>>>>
>>>>> ...
>>>>> run("Trainable Weka Segmentation");
>>>>> call("trainableSegmentation.Weka_Segmentation.loadClassifier",
>>>>> "/path/classifier.model");
>>>>> call("trainableSegmentation.Weka_Segmentation.getProbability");
>>>>> selectWindow("Probability maps");
>>>>> ...
>>>>>
>>>>> The problem is that fiji doesn't seem to want to wait for the
>>>>> classification. I tried inserting "wait(100000)" before
>>>>> "selectWindow("Probability maps"), but it didn't help and I don't
>> think
>>>>> there was even a wait.
>>>>>
>>>>> Any help would be much appreciated!
>>>>> Will
>>>>>
>>>>> --
>>>>> ImageJ mailing list:http://imagej.nih.gov/ij/list.html
>>>>>
>>>>>   I had a similar problem with the EDF plugin. Here is a quote form
>> that
>>>> post:
>>>>
>>>>> The problem I have is that the macro does not wait for EDF to finish,
>>> and
>>>>> EDF can take a while.
>>>>> My solution to this is based on the fact that EDF creates a window
>>> called
>>>>> "Output". so I did the following:
>>>>>
>>>>> initTime = getTime();
>>>>> while ( !isOpen("Output") ) {
>>>>>      elapsedTime = getTime() - initTime;
>>>>>      if (elapsedTime%10000 == 0) {
>>>>>          print(elapsedTime/1000, " seconds elapsed");
>>>>>          wait(10);
>>>>>      }
>>>>>
>>>>> which works (as  long as there is no other window called "Output"). I
>>>>> print an occasional reminder that the macro is still alive within this
>>> wait
>>>>> loop.
>>>>>
>>>> I posted this as a question, wondering if there is a better way to do
>> it.
>>>> I did not receive any reply, so perhaps in a macro this is the way to
>> go.
>>>> You should replace "Output" with "Probability maps".
>>>>
>>>> --aryeh
>>>>
>>>> --
>>>> Aryeh Weiss
>>>> Faculty of Engineering
>>>> Bar Ilan University
>>>> Ramat Gan 52900 Israel
>>>>
>>>> Ph:  972-3-5317638
>>>> FAX: [phone]
>>>>
>>>>
>>>>
>>>> --
>>>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>>>
>>>> --
>>>> Ignacio Arganda-Carreras, Ph.D.
>>>> Seung's lab, 46-5065
>>>> Department of Brain and Cognitive Sciences
>>>> Massachusetts Institute of Technology
>>>> 43 Vassar St.
>>>> Cambridge, MA 02139
>>>> USA
>>>>
>>>> Phone: [phone]
>>>> Website: <http://imagej.nih.gov/ij/list.html>
>>>> http://bioweb.cnb.csic.es/~iarganda/index_EN.html
>>>>
>>> --
>>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>>
>> --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
Aryeh Weiss
Faculty of Engineering
Bar Ilan University
Ramat Gan 52900 Israel

Ph:  972-3-5317638
FAX: [phone]

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html